Source Apportionment of Fine Organic Carbon at an Urban Site of 1 Beijing using a Chemical Mass Balance Model 2

21 22 Fine particles were sampled from 9 th November to 11 th December 2016 and 22 nd May 23 to 24 th June 2017 as part of the Atmospheric Pollution and Human Health in a Chinese 24 megacity (APHH-China) field campaigns in urban Beijing, China. Inorganic ions, trace 25 elements, OC, EC, and organic compounds including biomarkers, hopanes, PAHs, n- 26 alkanes and fatty acids, were determined for source

and offline data, differing at the temperature and relative humidity as specified above.  Table   239 1.  303 The composition of PM2.5 applying the chemical mass closure method is plotted in Fig.2 304 and summarized in Table S1. Because the gravimetrically measured mass (offline PM2.5) 305 differs slightly from online PM2.5 (Fig. S2), the regression analysis results between mass 306 reconstructed using mass closure (reconstructed PM2.5) and both measured PM2.5 307 (offline PM2.5/ online PM2.5) were investigated and plotted in Fig. 3.  and R 2 approaching unity (Fig. S3). This could indicate some uncertainties in offline 323 and/or online PM2.5 measurement for heavily polluted samples, or the applied OM/OC 324 ratio in winter was not suitable for converting OC to OM in heavily polluted samples.

325
During the summer campaign, the slope of the reconstructed PM2.5 and online-PM2.5 326 was close to 1, but that of reconstructed PM2.5 and offline-PM2.5 was 1.26. This could 327 be due to the loss of semi-volatile compounds from PTFE filters or the positive artefacts 328 of quartz filters for chemical analyses, which can absorb more organics than PTFE 329 filters that are used for PM weighing. To avoid loss of semi-volatiles, all collected 330 samples were stored in cold conditions, including during shipment. The datapoints were 331 more scattered in summer, which could result from the large difference in OM-OC 332 relationships from day to day. The reconstructed inorganics (reconstructed PM2.5 333 excluding OM) correlated well with offline-PM2.5, but OM did not (Fig. S4). Hence, 334 the discrepancies of between reconstructed PM2.5 and offline/online PM2.5 in summer 335 may be mainly attributable to variable OM/OC ratios.

340
Bound water contributed 4.6% and 7.2% of PM2.5 during the winter and summer, 341 respectively. All other components combined accounted for 13.2% and 12.4% of PM2.5 342 during the winter and summer campaigns, respectively.   that of rural household coal use (5% of decrease) (Zhao et al., 2018). In this study, coal 377 combustion is the single largest source that contributed to primary OC in both winter 378 and summer. In addition, industrial CC was a more significant source of OC than 379 residential CC in urban Beijing. On average, coal combustion related OC was 7.5±5.0 380 µg m -3 (34.5±9.8% of OC) in winter, which was more than 3 times of that in summer - in winter compared to that in summer. On haze days, industrial CC and residential CC 388 derived OC were 3.6 and 3.1 times that on non-haze days, respectively, indicating an 389 important contribution to haze formation from industrial CC.

390
Coal combustion is also a major source for particulate chloride (Chen et al., 2014).

391
Because Beijing is an inland city, the contribution of marine aerosols to particulate Cl -392 is considered minor, which is also supported by the higher Cl − /Na + mass ratios in winter

511
Even though the Other OC concentration was lower in summer, its relative 512 abundance was higher than that in winter, suggesting relatively higher efficiency of 513 SOA formation in summer due to more active photochemical processes under higher 514 temperature and strong radiation. The Other OC on winter haze days was 7.4±5.6 µg 515 m -3 , approximately 3 times of that on non-haze days (2.5±1.4 µg m -3 ). Other OC is also 516 compared with the SOC estimated by EC-tracer method below.

557
The OC source apportionment results in this study are also compared with those in results is presented in Fig. 7 and Table S3.
As shown in Fig. 7 and observations. Cooking accounted for over 10% of OC at the urban site, but less than 5% 586 at the rural site, which is plausible as the urban site is more densely populated.     AMS is plotted in Fig. 9. A similar temporal trend was found between them, especially 686 in summer, which was also observed with a better correlation (R 2 =0.73).   The source contributions to PM2.5 were calculated by multiplication of the fine OC 702 source estimates from CMB by the ratios of fine OC to PM2.5 mass (Table S4) Table S5. 724 As shown in Table S5, PM2.5 mass was well explained by those sources which 725 accounted for 91.9±24.1% and 99.0±19.1% of online PM2.5 in winter and summer, 726 respectively. In the summer, the offline PM2.5 is lower than online observations. Thus, 727 the CMB-based source contributions are more than offline PM2.5 mass (121.7±26.6%).

728
On average, the source contributions in winter ranked as SNA > coal combustion >